top of page

Why Business Intelligence implementations fails.

Roman Echeverria
Jun 29, 2021
2 min read

Updated: Jul 8, 2021


While building a BI platform many people focus on the visualizations, and while this is important we also need to build solid a background structure, we need to have a solid integration (ETL) foundation and a well designed data warehouse to store our data so we can access it efficiently. If you are leading the implementation make sure your team is composed by integration (ETL), data warehouse and business intelligence experts. It also important to consider business analyst to understand what actually the business needs, and finally to consider a designer if you are wishing to have not only functional but a esthetic designed platform.




Integration (ETL) Layer Consider wisely the information that your BI dashboards will need, then identify the systems/sources you will need to access, extract only the data that you need (on Data Society we are believers of the minimalism, this will let you build lightweigth programs, fast and easy to mantain and develop). We encorage you to use delta load mechanisms to pull only the most recent data, and avoid having full refreshes that will become slow over time. Dataware house Layer Create the data structures that you need to store the source system data, as a first step we encorage you to load the data into stage tables, that will store raw data from the source systems . Then you can use that data to build datamarts using any tranformations that you need. Don't forget creating indexes and table partitions whenever is needed to improve data access performance. Business Intelligence Layer During my years of experience as a consultant I have seen BI platforms with hundreds of KPIS and reports that are rarely used by the owners, I suggest to focus on what actually adds value to the user/customer and follow a minimalism approach. Gather with the stake holders identify the KPI's that add value to the business, and build reports and dashboards accordingly, organize them on a workspace that can be easily accessed, don't overwhelm the user with data, keep it simple and useful. Whenever possible refresh the dashboards during midnight so the data will be precomputed providing faster results to the user. If the user needs real time access to the data, make sure the underlaying data is well indexed and partitioned to avoid slowness problems.


 
 
 

Comments


Data Creed - Black Background.png

©2027 Data Creed

bottom of page